Multi-Model Fusion Harvested Energy Prediction Method for Energy Harvesting WSN Node

This paper proposes a precise and low overhead harvested energy prediction method for energy harvesting Wireless Sensor Network nodes, which is important for proper power management. This method merges multiple prediction models' results to achieve higher accuracy. A scheme for solar radiation prediction was implemented. Experiments using real solar data show that it improves the prediction error for all prediction intervals and lowers battery capacity required by more than 26.4%. It can run on light-weight MCUs, e.g. MSP430.